Performance of cluster plotter with large datasets and timelapse data

[stefanhahmann]

• When using the cluster plotter on large datasets, e.g. timelapses with >500 frames and >500 objects per frames, the cluster plotter is not very fluent anymore when playing the timelapse-movie.

• The reason for this is that the cluster diagram is repainted with all (in this case 500*500 = 250.000) points at each point of the timelapse, making the timelapse video very slow for these cases. This is hindering explorative data science on such datasets since the UI feels to slow to the user.

Potential solutions:

1. Run the painting of the diagram in a separate thread
2. Stop painting all points, once the timelapse moves on to the next frame
3. Only paint a subset (e.g.) 10% of the points if the dataset becomes too large (e.g. larger than 100.000 points)
4. Use an OpenGL supported plot library that can deal better with painting such an amount of points